# Harvard Dressed Its Qubits in Sound, and They Held Their Memory Three Times Longer By Kuraish Hosen · Science · Published Mon, 14 Sep 2026 04:02:41 GMT · Updated Mon, 14 Sep 2026 08:03:50 GMT Source: The Current Tribune — https://currenttribune.com/article/harvard-dressed-qubits-phonons-silicon-vacancy-diamond-nature-physics Every quantum computing story eventually arrives at the same villain. A qubit holds information beautifully and briefly, and then the world leans on it — a stray magnetic field, a temperature wobble, the low-frequency hum of everything nearby — and the information is gone. Coherence time is the whole ballgame. Extend it and the machine works. Fail to, and you are building an extremely cold random number generator. A team at Harvard’s School of Engineering and Applied Sciences has now tripled the coherence time of a particular kind of qubit using a tool that sounds wrong for the job: sound itself. The work, published in *Nature Physics*, comes out of Marko Lončar’s lab, with experiments led by Eliza Cornell — who has since moved to a postdoc at Boston University — and former postdoc Zhujing Xu. ## The Trick: Dress the Qubit Instead of Isolating It The standard defense against decoherence is a technique called dynamical decoupling. You hit the qubit with a carefully timed sequence of microwave pulses that average out the noise, a bit like rotating a spinning top to cancel a wobble. It works. It also requires bulky microwave hardware sitting next to every qubit, which is a problem when your goal is to fit thousands of them on a chip. The Harvard approach drops the pulses entirely. Instead of hitting the qubit intermittently, the team applies a continuous mechanical driving field — a steady acoustic vibration — to a silicon-vacancy spin center inside a diamond crystal. That constant drive pushes the qubit into what physicists call a dressed state. The qubit is no longer a bare spin exposed to its surroundings; it is a spin wearing an acoustic field, and the combination has different noise sensitivities than the spin alone. Specifically, it becomes far less responsive to the low-frequency noise that does most of the damage to quantum memory. Cornell’s framing of the underlying tension is the clearest way to understand why this is clever: “We are solving two problems. We want the spin to have strong interaction with phonons, and we want the spin to have a long coherence time.” Those two goals normally fight each other. Anything a qubit couples to strongly is also a channel through which noise can reach it. Coupling tightly to phonons should make the qubit more fragile, not less. The dressing is what flips the sign — the same mechanical field that carries information also becomes the shield. ### Why phonons and not photons Most quantum networking proposals use light. Photons are fast, they travel well, and the optical toolbox is mature. They also have a practical disadvantage nobody can engineer away: wavelength. A phonon of a given frequency has a wavelength orders of magnitude shorter than a photon of the same frequency, because sound moves through material vastly slower than light moves through vacuum. Shorter wavelength means smaller structures. A phononic cavity that confines a vibration can be dramatically more compact than the optical equivalent, which is the difference between a quantum network you build on a chip and a quantum network you build on an optical table. The Harvard result matters because the protection mechanism is built from the same physics as the interconnect. The phonons that would shuttle information between stationary nodes in a future network are the phonons doing the shielding. You are not bolting a protection scheme onto a communication scheme; they are the same structure. ## What Silicon-Vacancy Centers Bring The qubit here is a silicon-vacancy center — a defect in the diamond lattice where a silicon atom sits between two missing carbon atoms. These defects have become a favorite for quantum networking because they combine a usable spin with clean optical properties, and diamond is a rigid, low-loss host for acoustic waves. They have historically had a coherence problem, which is precisely the problem this paper addresses. - **Result:** roughly threefold extension of silicon-vacancy spin coherence time - **Method:** continuous mechanical driving to create a dressed state, no microwave pulse sequences - **Host material:** diamond crystal with silicon-vacancy defect centers - **Key advantage:** compatible with the phononic cavities already designed for quantum network nodes The author list is a reasonable map of who is working on this problem: alongside Cornell and Xu, the paper includes Zhaoyou Wang, Hana K. Warner, Eliana Mann, Michael Haas, Smarak Maity, Graham Joe, Liang Jiang, Peter Rabl and Benjamin Pingault. ## The Honest Limits Threefold is a real improvement and it is not a solved problem. Fault-tolerant quantum computing needs coherence times measured against gate operation times with enormous margin, and the gap between where solid-state spin qubits sit today and where error correction becomes practical is still large enough to require several more results like this one. Continuous driving also has costs that any implementation has to pay. A steady mechanical field means steady power going into the system, and power means heat, which is an unwelcome guest in a dilution refrigerator. The technique buys protection against low-frequency noise specifically; it does not offer blanket immunity, and noise at the driving frequency itself becomes a new thing to worry about. And this is one device demonstrating a principle. The distance between a working physics experiment and a manufacturable array of thousands of identical devices is, historically, where most quantum computing ideas go to die. ## What This Means The significance here is architectural rather than numerical. Three times better coherence is a good result; the reason to pay attention is that it arrives without adding hardware. Quantum computing’s scaling problem is not purely about qubit quality. It is about the enormous supporting apparatus each qubit demands — control lines, microwave sources, filtering, wiring that has to reach into a cryostat without carrying heat. Every technique that folds a function into the structure already present is worth more than its raw numbers suggest, because it removes a component that would otherwise have to be replicated thousands of times. The longer-term picture the Lončar lab is sketching is a hybrid one: chips where acoustic waves move quantum information between nodes, protect it in transit and in storage, and interface with other qubit types that are better suited to computation than communication. That is a more plausible near-term architecture than a single monolithic quantum processor, and it is the direction a lot of serious hardware work has quietly been heading. None of this puts a quantum computer in a data center next year. What it does is remove one more reason to think the phonon-based approach is a dead end — and in a field where most promising directions eventually reveal a fatal scaling limit, surviving another test is how progress actually looks.